A nonprofit's weekday AI task drafts its support inbox replies

You can have an AI tool read a support inbox every weekday morning and leave draft replies for you to check, built from documents you already maintain. Vegan Hacktivists does this for Amplify, "our AI training program for animal advocates," and wrote up the setup in a guide updated Oct 27.

What it replaces#

Participants email about "registration, session times, recordings, and access." The guide sums up the change: "That gives us replies to review without looking up the same course details each time."

How it works#

"We set up a scheduled task that runs each weekday." For support email, the guide says it:

1. "Reads new support emails and the full conversations, including previous replies and existing drafts."
2. "Looks up the answer in our course documents or registration records."
3. "Inserts draft replies for us to review and flags questions that need our input."
4. "Archives routine notifications and resolved conversations under rules we agreed on."

Their prompt is published in full. It starts "I want your help managing our support inbox," asks the AI to "Interview me about our inbox, reference documents, reply style, archive rules, and how I want results presented," and ends "Keep replies as drafts for review, and return one linked bullet per item."

It is in regular use. Steven Rouk, Director of AI at Vegan Hacktivists per his X bio, wrote on Oct 29 that "I personally have these running." One is a daily chief of staff "scheduled task in ChatGPT (desktop app) that surfaces important emails for me each morning". The other is an Amplify operations chat that "inserts drafts to participant emails where it can."

Good to know#

• Results are self-reported. The guide gives no counts or time saved.
• The guide lists "Claude, ChatGPT, Gmail" but doesn't say which tool runs the support task. Steven's post places only the chief-of-staff task in ChatGPT.
• Test before you schedule. The guide says "access varies by app, plan, and workspace settings. Ask it to flag anything it cannot check."
• Archiving is the risky part. Theirs archives only "under rules we agreed on." Leave it off until your rules are written down.
• Disclosure: the guide is published by Steven's employer, a nonprofit that "offers free AI and automation support to animal protection organizations," and the site carries a newsletter signup.

Set it up in Grok Bot#

A Grok Bot routine with Gmail connected can play the same role. Cursor's routines page notes "Each run spends usage," so a once-a-weekday run is cheap next to an hourly one.

Starter prompt (ours, untested):

Every weekday at 8 AM, read new emails in my support inbox and the
full thread for each, including my earlier replies and any drafts.
Answer only from these documents: [links to FAQ, schedule, records].
For each email you can answer, save a reply draft in the thread.
List the ones that need me, with one linked line each and why.
Don't send, archive or delete anything. Before the first run,
interview me about my reply style and what counts as urgent.

For a paid-product version with refunds and approvals, see our support playbook.

Last checked: Nov 2, 2026, 7:55 to 7:57 AM ET. What changed: first published Nov 2.